In this fifth blog in the “Keeping the light on: Reflections on GEI and AI in Africa” series, Dr Elizabeth Oseku, project coordinator for the Hub for Artificial Intelligence in Maternal, Sexual and Reproductive Health in Sub-Saharan Africa(HASH), part of AI4D, reflects on her personal and hub’s journey to put herself in the shoes of the diversity of people with sexual and reproductive health needs. She thinks about the importance of more intersectional data in AI driven solutions.
Girls can do what boys can do
Our project, the Hub for Artificial Intelligence in Maternal, Sexual Reproductive Health, started in November 2021 with a multidisciplinary team of medical doctors, social scientists, computer scientists and data scientists. As a team, the majority of us had a general idea about what gender, equity and inclusion meant. For most of us, it meant affirmative action to ensure representation of women and achieve gender balance.
We were sure that as a team we were at least doing well in terms of that; after all, our Lead Principal Investigator is a woman! And in case there was anything else we had missed, one of our computer scientists (a woman, by the way) was a gender expert and would cover that.
As an individual, my own ideas around GEI were not much more than believing that girls can do what boys can do. So as long as girls can have an education and rewarding work, I did not have much more to say. I grew up in an environment that never stifled my ambitions, therefore I felt that every girl should grow up with that liberty.
We saw things in a way we had never seen before
As the project progressed, AI4D and IDRC organised various means to learn more about the concept of GEI. They shared reading material, they organised webinars and assigned us our very own gender expert from Gender At Work , Marie-Katherine (Kate) Waller. Kate planned regular catch-ups with us and helped us explore our perceptions about the concept.
As we listened to each other, we saw things in a way we had never seen before. We saw differences between and within genders. We each gained a better understanding of how differences in gender impacted the freedom, competence and confidence with which people made choices about their own bodies, their futures and their health. We also saw that within the same gender, the path to exercising one’s full rights is not always linear and depends on one’s background and exposure.
These realisations introduced to us the concept of intersectionality. I am compelled to think, for example, about the four principles of medical ethics that are meant to guide medical practice – autonomy, beneficence, justice and non-maleficence. These too have historically been exercised more or less, depending on the complex factors of intersectionality such as sex, religion, race, level of education, age, ethnicity,etc.
Regarding AI for health, we saw that from a GEI perspective, contrary to what is usually the case, in our area of maternal, sexual and reproductive health (MSRH), it is usually men that are under represented in the data and therefore there is more likely to be a bias towards development of AI tools that benefit women more than men.
One of our male data scientists, Wisdom Favor Ceasar, was so taken by GEI that when Kate suggested that we select a specific group of people from our team to champion GEI for our project, he offered to lead it. As an AI innovator, he felt that the AI field needed a framework for thinking about GEI in AI for MSRH given the multiple factors associated. Therefore our team set out to develop this. Our team also started to represent HASH at the Peer Learning Journey sessions where we shared our experiences with other AI4D members.
When we onboarded our ten subgrantees, we realised that they were in a similar place to where we had been at the start of our project in terms of GEI. Having now grown a little bit ourselves, we knew what to do. We knew the value of knowledge resources, of community and of mentorship, and so we set up mechanisms to replicate these. Since then, we set up a GEI club that runs regular GEI focused sessions and we are also working on our framework to guide incorporating GEI into a MSRH AI project. We also perform quarterly check-ins with our subgrantees where GEI is emphasised. In these check-ins our GEI team champions were keen to probe about how the GEI can be further incorporated into the projects at the different project stages, e.g. data collection, data analysis, user testing, etc.
Going forward
As our project draws to a close, I have learned that one of the keys to practising GEI in AI is putting yourself in the shoes of another. Only then do concepts like ethics, intersectionality and bias come to life. One of the eureka moments for me was realising that the accuracy of an AI tool that predicts miscarriages, can be improved by adding data around social factors like intimate partner violence; and yet in the usual clinical datasets a developer might think to collect, this may not be present. It would take imagining the lives of different mothers who come to a clinic in order to plan for this kind of dataset. There are so many things we never see or think about when we only see things through our own myopic lens. It has been quite a journey. I now have more to say but there is still more to learn.
Dr Elizabeth Oseku, MBChB, MSc PH is a Medical Doctor, Public Health specialist and project manager at the Infectious Diseases Institute in Uganda. She currently coordinates the research Hub for Artificial Intelligence in Maternal, Sexual and Reproductive Health in Sub-Saharan Africa (HASH). Her research interests are around digital health (ranging from Health to Artificial Intelligence), health promotion and the wider determinants of health. You can find her on LinkedIn.
Stay tuned for the next blog posts in the series Keeping the light on: Reflections on GEI and AI in Africa!